Lefschetz hyperplane theorem

Solomon Lefschetz[3] used his idea of a Lefschetz pencil to prove the theorem. Rather than considering the hyperplane section Y alone, he put it into a family of hyperplane sections Y_t, where Y = Y₀. Because a generic hyperplane section is smooth, all but a finite number of Y_t are smooth varieties. After removing these points from the t-plane and making an additional finite number of slits, the resulting family of hyperplane sections is topologically trivial. That is, it is a product of a generic Y_t with an open subset of the t-plane. X, therefore, can be understood if one understands how hyperplane sections are identified across the slits and at the singular points. Away from the singular points, the identification can be described inductively. At the singular points, the Morse lemma implies that there is a choice of coordinate system for X of a particularly simple form. This coordinate system can be used to prove the theorem directly.[4]

Aldo Andreotti and Theodore Frankel[5] recognized that Lefschetz's theorem could be recast using Morse theory.[6] Here the parameter t plays the role of a Morse function. The basic tool in this approach is the Andreotti–Frankel theorem, which states that a complex affine variety of complex dimension n (and thus real dimension 2n) has the homotopy type of a CW-complex of (real) dimension n. This implies that the relative homology groups of Y in X are trivial in degree less than n. The long exact sequence of relative homology then gives the theorem.

Neither Lefschetz's proof nor Andreotti and Frankel's proof directly imply the Lefschetz hyperplane theorem for homotopy groups. An approach that does was found by René Thom no later than 1957 and was simplified and published by Raoul Bott in 1959.[7] Thom and Bott interpret Y as the vanishing locus in X of a section of a line bundle. An application of Morse theory to this section implies that X can be constructed from Y by adjoining cells of dimension n or more. From this, it follows that the relative homology and homotopy groups of Y in X are concentrated in degrees n and higher, which yields the theorem.

Kunihiko Kodaira and Donald C. Spencer found that under certain restrictions, it is possible to prove a Lefschetz-type theorem for the Hodge groups H^p,q. Specifically, assume that Y is smooth and that the line bundle ${\displaystyle {\mathcal {O}}_{X}(Y)}$ is ample. Then the restriction map H^p,q(X) → H^p,q(Y) is an isomorphism if p + q < n − 1 and is injective if p + q = n − 1.[8][9] By Hodge theory, these cohomology groups are equal to the sheaf cohomology groups ${\displaystyle H^{q}(X,\textstyle \bigwedge ^{p}\Omega _{X})}$ and ${\displaystyle H^{q}(Y,\textstyle \bigwedge ^{p}\Omega _{Y})}$. Therefore, the theorem follows from applying the Akizuki–Nakano vanishing theorem to ${\displaystyle H^{q}(X,\textstyle \bigwedge ^{p}\Omega _{X}|_{Y})}$ and using a long exact sequence.

Combining this proof with the universal coefficient theorem nearly yields the usual Lefschetz theorem for cohomology with coefficients in any field of characteristic zero. It is, however, slightly weaker because of the additional assumptions on Y.

Michael Artin and Alexander Grothendieck found a generalization of the Lefschetz hyperplane theorem to the case where the coefficients of the cohomology lie not in a field but instead in a constructible sheaf. They prove that for a constructible sheaf F on an affine variety U, the cohomology groups ${\displaystyle H^{k}(U,F)}$ vanish whenever ${\displaystyle k>n}$.[10]